Why Bad Data Is the Biggest Barrier to AI Adoption in Manufacturing

For manufacturers trying to figure out where to start with AI adoption in manufacturing, the answer is almost never the AI itself. It’s the data behind it.

That’s one of the clearest takeaways from a recent Moore on Manufacturing episode featuring Daniel Shorstein, president of digital services at James Moore & Company, alongside hosts Mike Sibley and Kevin Golden. The conversation cut through a lot of the noise around AI to focus on something more fundamental: if your data isn’t reliable, no amount of automation or artificial intelligence is going to fix your business.

“My Data Stinks” Is More Common Than You Think

Shorstein works across industries, from federal government to higher education to manufacturing, and he hears the same thing everywhere.

“I can’t even automate anything. I can’t even use AI because my data stinks. I can’t rely on that,” he said, describing what clients tell him when they first explore automation or AI.

It sounds like a simple problem, but the reasons behind it are usually layered. Data comes from multiple systems, including ERP platforms, equipment, CRM tools and financial software, and those systems weren’t built to talk to each other cleanly. On top of that, the definitions behind the numbers are often inconsistent.

Shorstein gave a straightforward example: “You look at a dashboard that says sales and then you look at another report that says sales and you’re like, why don’t these numbers tie? Oh, well, that one’s net of discounts and that one excludes e-commerce.”

If two people in the same organization are looking at the same word and seeing different numbers, you don’t have a data problem. You have a trust problem. And trust is exactly what AI depends on.

Why Data Quality Comes Before AI Strategy

A lot of manufacturers jump into AI conversations at the wrong level. They’re asking which tool to use before they’ve confirmed whether their underlying data is accurate enough to feed into that tool.

Shorstein’s recommendation is to start by asking a different question entirely: what do you actually need to know?

“What are the things that you need to know that you don’t know today? What questions are you trying to answer that you can’t?” he said. “Use that as the driver rather than just saying, let me collect all of my data and create a bunch of charts because that would be cool.”

Once you know the questions, you can work backward to figure out what data you need to answer them. Then you can build a strategy around collecting, cleaning and centralizing that data in a way that’s testable and reliable. Jumping straight to dashboards or automation before that work is done is how manufacturers end up with fast pipelines delivering wrong information.

“If you build it too fast, you might end up in a situation where you’re getting information fast, but if it’s the wrong information, it’s going to cause problems,” Shorstein said. “I think my margin’s 40% but it ends up only being 20% because I forgot to pull that one field in.”

Sibley echoed this from the client side: “Kevin and I have both seen lately with some prospective clients who have made bad decisions based on bad information that they didn’t realize was bad. That’s a killer for a business.”

The Documentation Gap Nobody Talks About

Even when manufacturers have data, they often lack documentation around what that data actually means. Who defined the formula? Which fields were included? Were discounts netted out or not?

Without that documentation, every person who builds a report or dashboard is making their own interpretation. Over time, those interpretations drift, and you end up with a patchwork of numbers that nobody fully trusts and nobody can fully explain.

Shorstein’s advice is to get the definitions down before building anything. “It’s important to have some data strategy and some architecture and some documentation, so that way whoever is trying to build these things for you, you have some consistency and some documentation around the definitions of what you’re even trying to look at.”

This isn’t glamorous work. But it’s the work that makes everything else possible.

Start Small, Validate Heavily, Then Build

The natural instinct when building a data and automation strategy is to go broad. Pull everything. Connect all the systems. Build the full dashboard. Shorstein pushes back on that approach consistently.

“Start simple,” he said. “Start with one or two pieces of information that you can test heavily and test for a period of time, and just get very comfortable and confident that it is the right information.”

Once you’ve validated that a data source is reliable and that the numbers it produces can be trusted, you can build on top of it. Add another source. Automate the reporting. Layer in AI where it makes sense. But each step should be tested before you move to the next one.

This is where the rules-based automation conversation becomes relevant. Before AI enters the picture, there’s often an opportunity to automate data collection and reporting through traditional methods. That foundational work makes any future AI application more accurate and more useful.

What This Means for Manufacturers Right Now

If your organization is thinking about AI and you haven’t yet addressed your data quality, you’re not behind on AI. You’re behind on data. That’s actually a more solvable problem, and solving it creates the foundation for everything else.

The sequence Shorstein recommends: understand what questions you’re trying to answer, identify what data you need to answer them, clean and document that data, and then start building reliable pipelines that surface the right information at the right time.

After that, the conversation about automation and AI becomes a lot more productive.

To hear the full discussion, including how manufacturers can apply AI to strategic planning, margin analysis and workflow improvement, watch the complete Moore on Manufacturing episode.

All content provided in this article is for informational purposes only. Matters discussed in this article are subject to change. For up-to-date information on this subject please contact a James Moore professional. James Moore will not be held responsible for any claim, loss, damage or inconvenience caused as a result of any information within these pages or any information accessed through this site.